This thesis is concerned with the problem of optimally designing step-stress and constant stress accelerated life test(ALT) plans for products with multiple modes of failure. The model consists of independent exponential life distributions with mean that is a log-linear function of stress for each failure mode, and a cumulative exposure model for the effect of changing stress.
Minimizing the sum of asymptotic variances of the maximum likelihood estimators of the mean lives at design condition for all failure modes is used as an optimality criterion.
In step stress ALTs, three cases where the test procedure is observed continuously in time until a prespecified censoring time are considered;
(1) Two stress levels are used and stress is changed to high level when a specified time is run at the low stress. (Simple time-step ALT)
(2) Two stress levels are used and stress is changed to high level when a specified propotion of units fail at the low stress. (Simple failure-step ALT)
(3) Multiple stress levels are used and stress is changed at every specified constant time interval. (Multiple time-step stress ALT)
For each case, optimal stress change point is obtained and its behaviors are analyzed.
In constant stress ALT, two stress levels are used and the high stress level is assumed to be given. The optimal low stress level and the proportion of units tested at each stress are obtained and their behaviors are analyzed, and tables for finding optimum plans are given.